Thema: "Developing a knowledge acquisition and evaluation method for a chatbot"
Chatbots capable of natural-language interaction are finding uses in a wide range of applications, such as undergraduate advisor, education and, notably, customer service, allowing to promptly answer customer’s inquiries.
An important feature of a chatbot is robustness understood as ability to handle the inputs that differ from predefined patterns, either because of mis- spellings or different wording. From the point of view of an institution using a chatbot, another desirable feature is ease of maintenance understood as ability to add new or edit interaction scenarios easily.
This thesis is being done in cooperation with the Novomind company, which already utilises a rule-based chatbot to handle inquiries of potential customers, and looks for the development of a chatbot that would be more robust and flexible than the current regex-based solution. Their existing solution requires exact match of the user input to a predefined regular expression and is hard to maintain due to a large number of existing rules, and these issues need to be addressed by proposing a consistent methodology for chatbot development using state-of-the-art tools, with maintenance and scalability in mind.